fix: isolate relay workers by desktop keyring session - #75
Conversation
🦋 Changeset detectedLatest commit: 8f6f7e6 The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c900c4f89a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "DBUS_SESSION_BUS_ADDRESS", | ||
| "DISPLAY", |
There was a problem hiding this comment.
Adding these entries changes the serialized profile input even when both variables are unset, so every existing 0.6.0 relay receives a new profile after upgrading. With the default location, commands start against an empty directory while the old detached worker and its routes, checkpoints, and pending records remain active but inaccessible; with GROK_BOT_RELAY_DIR, startup instead permanently rejects the existing ledger with Relay profile mismatch. Preserve or migrate the previous profile/state so upgrades continue existing durable routes rather than stranding them.
Useful? React with 👍 / 👎.
| "DBUS_SESSION_BUS_ADDRESS", | ||
| "DISPLAY", |
There was a problem hiding this comment.
These variables are hashed on macOS too, although readKeychainPassword() uses /usr/bin/security there rather than Secret Service. Consequently, otherwise compatible macOS callers that differ only because one runs under XQuartz, SSH forwarding, or a Homebrew D-Bus session get separate state directories and workers, making existing routes disappear from status and permitting duplicate bindings. Include these fields only for the Linux keyring path.
Useful? React with 👍 / 👎.
The windows-latest runs were cancelled while still queued and never executed. This token cannot rerun Actions, so push a no-op commit to start new checks. Co-authored-by: Zack Jackson <ScriptedAlchemy@users.noreply.github.com>
2 participants
Automatic replies reused a headless MCP worker whose environment could not unlock the desktop keyring, while ordinary CLI sends worked. Include the D-Bus session address and DISPLAY in the worker profile so those callers cannot share incompatible credential environments.
Validation: reproduced identical profiles before the fix; regression passes afterward. 31 focused worker, app-session, and auth-recovery tests pass. Includes patch changeset. This fixes worker reuse; direct headless MCP credential access remains a separate host configuration requirement.